The gene sets reported here represent significant additions to the pool of identified olfactory genes in Coleoptera. Prior to this study, members of the major chemosensory gene families in Coleoptera had been identified only from the genome of T. castaneum (except for the ORs also identified in the antennal transcriptome of the cerambycid, M. caryae). Additionally, as the genes identified here underlie the aggregation behavior that results in tree killing by mass-attack, they represent novel targets for management programs of two of the world’s most destructive forest pests.
In general, we identified somewhat larger numbers of transcripts encoding putative olfactory proteins (i.e. ORs, IRs, OBPs, and CSPs) in D. ponderosae than in I. typographus. The greater depth of the 454-sequencing and the access to Sanger data for D. ponderosae likely account for this difference. In addition, duplex-specific nuclease cDNA normalization (performed only for I. typographus) seems to result in overrepresentation of shorter full-length transcripts, which might explain the lower number of OR and IR transcripts identified in I. typographus, and also the absence of Orco transcripts in the transcriptome assembly. However, despite the slight difference in methodology, the GO annotation demonstrated a remarkable overall similarity in the types of genes (with respect to associated GO terms) that are expressed in the antennae of the two species. GO annotation was previously conducted for the antennal transcriptome of Manduca sexta moths by Grosse-Wilde et al. , and comparison with their data reveals a striking similarity to the bark beetles analyzed here. Indeed, several GO terms with putative relation to olfactory function showed identical (or near identical) relative abundance, suggesting a kind of across-order conservation of gene expression patterns in antennae, although the data say nothing about expression levels of the individual genes themselves.
Odorants are thought to interact with OBPs or CSPs in the sensillum lymph prior to the ligand-receptor interaction. The numbers of OBPs identified in the bark beetles (15 in I. typographus and 31 in D. ponderosae) are clearly lower than the 49 OBP-encoding genes reported in the genome of T. castaneum. The same is true for the CSPs, for which we identified 6 transcripts in I. typographus and 11 in D. ponderosae compared with 20 putative CSP encoding genes in the T. castaneum genome . However, it might be misleading to compare the number of genes identified in a genome with the number of transcripts in a specific tissue at a specific life stage. Some of the genes might, for instance, be expressed only in the larva [53, 67]. Indeed, many of the identified OBPs and CSPs (ca. one third of the transcripts) in D. ponderosae were not identified from the antennal library, but seem to be expressed only in non-olfactory tissue. Similar patterns have been found also in other insects, suggesting that these proteins may have physiological functions independent of olfaction [41, 45, 68].
SNMPs are associated with pheromone-responsive OSNs in Lepidoptera and Diptera [50, 51]. In D. melanogaster, SNMP1 was shown to be necessary for proper OSN responses to the pheromone compound cis-vaccenyl acetate, but not for OSN responses to food-related fruit esters . Benton et al.  also demonstrated that SNMP was required for activation of Heliothis virescens (Lepidoptera) pheromone receptor HR13 by its corresponding ligand when heterologously expressed in Drosophila neurons. It was suggested that the hydrophobic tail of the fatty-acid derived dipteran and lepidopteran pheromone molecules necessitates the presence of SNMP. If so, that raises the question why bark beetles that do not use pheromone compounds with long hydrophobic tails  express SNMPs in their antennae.
The numbers of putative OR-encoding transcripts identified in the two bark beetles (43 in I. typographus and 49 in D. ponderosae) are close to the number reported in the antennal transcriptome of M. caryae (57 ORs) , but lower than the number expressed in the head of adult T. castaneum (111 ORs), and much lower than the number in the T. castaneum genome (341 OR-encoding genes, including 79 pseudogenes) . In other insects, the number of seemingly intact OR-encoding genes identified from genomes is highly variable , ranging from only 10 in the human body louse, Pediculus humanus, to ca. 300 in the fire ant, Solenopsis invicta. It is not fully understood how the number of ORs relates to the ecology of an insect. In our case, one could expect that the flour beetle might have a less complex sense of smell than the forest dwelling beetles, since it has presumably adapted to an environment with a lower “semiochemical diversity” . This would suggest a lower number of receptors, contrary to our results. Therefore, the chemical ecology of T. castaneum may be more complex than currently understood as also suggested by . However, it is unknown how many of the 111 ORs that are expressed in the adult head are actually expressed in the olfactory organs of T. castaneum. In addition, it is likely that some bark beetle ORs have been missed in our transcriptome analysis (especially in Ips due to the lower sequencing depth), underestimating the true number of antennal-expressed bark beetle ORs.
Species (or taxon)-specific expansions of OR lineages are seen in most insects studied e.g. [72, 73], and some of the largest expansions have been found in Hymenoptera, particularly in the jewel wasp, Nasonia vitripennis. The pattern of OR lineage expansion and conservation observed in the present study likely reflects the evolutionary and ecological relatedness among the four beetle species. The beetle taxa analysed here all belong to the more derived part of Coleoptera (Cucujiformia) . However, the Curculionidea (with Ips and Dendroctonus) and Tenebrionidea (with Tribolium) superfamilies are the two furthest separated clades within Cucujiformia, sharing a common ancestor ca. 230–240 Mya. Thus, it may come as no surprise that the ORs of these two taxa largely fall into different subgroupings in the tree. On the other hand, the Curculionidea is a sister group to the Chrysomeloidea (including the longhorns)  and, likewise, the closer relatedness of these taxa seems to be reflected in the OR subgroupings. Within Scolytinae, the Ips and Dendroctonus genera are separated by ca. 80 Mya . However, despite the fact that Culex and Aedes mosquitoes are separated by only ca. 40 Mya , they show more distinct species-specific OR lineage expansions than the bark beetles , indicating that ecological adaptation and life cycle also play important roles in shaping the OR repertoire of a species . On this note, the bark beetles and the cerambycid utilize similar types of host material, i.e. conifer trees and hardwood, respectively [2, 79], whereas T. castaneum has been associated with human populations and stored products, for at least a few thousand years .
However, not all ORs were grouped in taxon-specific expansions; some subfamilies contained ORs from all four species. This might indicate preservation of ancestral functional patterns within Coleoptera, but since non-coleopteran ORs were left out from the analysis we are careful to draw any conclusions based on this finding (i.e. the clades might contain receptors also from insects outside Coleoptera).
The close clustering of OR sequences from the two bark beetles raises the question about how similar the semiochemical environment is for I. typographus and D. ponderosae. They both live in conifers and would thus be expected to share several biologically relevant compounds. Due to their status as very serious forest pests, the plant- and beetle-produced compounds that they respond to are well studied in these two species. Mainly based on a set of review papers [2, 3, 7, 81–83], we compiled a table of all compounds that have been shown to be physiologically and/or behaviorally active in I. typographus and D. ponderosae (Additional file 5). For 29 (54%) of the 54 listed compounds, there is evidence of shared bio-activity. Not surprisingly, the host compounds show a large overlap (61%), but there is also a large overlap (56%) among pheromone compounds of beetle origin. For the non-host volatiles, the overlap is lower (40%). One might speculate that the extent of this shared “chemosphere” of semiochemicals could account for the low degree of species-specific diversifications among the bark beetle ORs and the other proteins studied here. However, functional data is required to test this hypothesis.
We identified only a small number of putative GR-encoding transcripts (6 in I. typographus; 2 in D. ponderosae) from the antennal transcriptomes. The identified bark beetle GRs included transcripts for carbon dioxide receptors, suggesting that the antennae of bark beetles detect carbon dioxide. In addition, the presence of GR1-3 in I. typographus indicates that carbon dioxide is detected by a heterotrimer receptor, like in mosquitoes, Bombyx mori, and T. castaneum[15, 84]. However, GR2 was not found in the analyzed transcriptome of D. ponderosae. Hence, it is possible that D. ponderosae uses a heterodimer receptor for carbon dioxide detection (like D. melanogaster) , but it seems unlikely that expression of GR2 would have been lost in only one of the bark beetle species analyzed here.
All the conserved antennal IRs that previously were found in T. castaneum were also identified in D. ponderosae. However, some of them were missing in the I. typographus data. As IRs are associated with coeloconic sensilla that are relatively rare on the Ips antenna , it is possible that the missing IR transcripts are expressed only in a few neurons. A lower expression level results in a higher probability that these transcripts were missed during the random sequencing of the Ips cDNA, which had a lesser depth than for D. ponderosae. Generally in insects, the antennal IR subfamily constitutes only a portion of the total number of IRs. The others belong to the divergent IRs, a subfamily that shows species-specific expansions that are particularly large in Diptera . In D. melanogaster, expression of divergent IRs was detected only in gustatory organs [9, 33]. This is consistent with the scarcity of divergent IRs in the bark beetle antennal transcriptomes.